Fix some #[cfg_attr(not(doc), repr(..))]

[GrigorenkoPV]

Now that #90435 seems to have been resolved.

[rustbot]

r? @tgross35

rustbot has assigned @tgross35.
They will have a look at your PR within the next two weeks and either review your PR or reassign to another reviewer.

Use r? to explicitly pick a reviewer

[tgross35]

The changes lgtm, I'll wait for CI artifacts to be available and take a look at those docs (mingw-check uploads API docs)

[tgross35]

Could you also update this bit of the docs?

rust/src/doc/rustdoc/src/advanced-features.md

Lines 115 to 129 in 92c6c03

	## `#[repr(transparent)]`: Documenting the transparent representation
	You can read more about `#[repr(transparent)]` itself in the [Rust Reference][repr-trans-ref] and
	in the [Rustonomicon][repr-trans-nomicon].
	Since this representation is only considered part of the public ABI if the single field with non-trivial
	size or alignment is public and if the documentation does not state otherwise, Rustdoc helpfully displays
	the attribute if and only if the non-1-ZST field is public or at least one field is public in case all
	fields are 1-ZST fields. The term *1-ZST* refers to types that are one-aligned and zero-sized.
	It would seem that one can manually hide the attribute with `#[cfg_attr(not(doc), repr(transparent))]`
	if one wishes to declare the representation as private even if the non-1-ZST field is public.
	However, due to [current limitations][cross-crate-cfg-doc], this method is not always guaranteed to work.
	Therefore, if you would like to do so, you should always write it down in prose independently of whether
	you use `cfg_attr` or not.

[tgross35]

Hm - this doesn't look like the docs changes applied. From the artifacts at https://github.com/rust-lang/rust/actions/runs/10031700259/artifacts/1723776160:

Vs. stable:

@fmease looks like you worked on this, is this supposed to be working now?

[tgross35]

@rustbot author

[fmease]

Could you also update this bit of the docs?

rust/src/doc/rustdoc/src/advanced-features.md

Lines 115 to 129 in 92c6c03

	## `#[repr(transparent)]`: Documenting the transparent representation
	You can read more about `#[repr(transparent)]` itself in the [Rust Reference][repr-trans-ref] and
	in the [Rustonomicon][repr-trans-nomicon].
	Since this representation is only considered part of the public ABI if the single field with non-trivial
	size or alignment is public and if the documentation does not state otherwise, Rustdoc helpfully displays
	the attribute if and only if the non-1-ZST field is public or at least one field is public in case all
	fields are 1-ZST fields. The term *1-ZST* refers to types that are one-aligned and zero-sized.
	It would seem that one can manually hide the attribute with `#[cfg_attr(not(doc), repr(transparent))]`
	if one wishes to declare the representation as private even if the non-1-ZST field is public.
	However, due to [current limitations][cross-crate-cfg-doc], this method is not always guaranteed to work.
	Therefore, if you would like to do so, you should always write it down in prose independently of whether
	you use `cfg_attr` or not.

I'm not entirely sure why those are in need of updating? I wrote them in #115439 and they 100% reflect how rustdoc behaves today.

[fmease]

Hm - this doesn't look like the docs changes applied. From the artifacts at https://github.com/rust-lang/rust/actions/runs/10031700259/artifacts/1723776160:

Vs. stable:

@fmease looks like you worked on this, is this supposed to be working now?

I don't quite know what the linked artifact represents because it seems like std got documented with --document-private-items there judging by CStr.inner being shown in the docs (instead of /* private fields */). If that's the case, it's no surprise (to me) that #[repr(transparent)] gets shown as I made rustdoc always show #[repr(transparent)] under --document-private-items (which makes sense to me).

[fmease]

I'd like to link dtolnay's attempt at doing something similar to this PR in #116743. However, it's closed because not only did that PR try to simplify #[repr(transparent)] attrs but any #[repr]s — which is still blocked on my PR #116882 (which itself is blocked — on design questions).

[GrigorenkoPV]

@rustbot r? fmease
@rustbot ready

[fmease]

[@]rustbot r? fmease

Libs is not within my purview, only compiler and rustdoc is.
r? tgross35

[tgross35]

I don't quite know what the linked artifact represents because it seems like std got documented with --document-private-items there judging by CStr.inner being shown in the docs (instead of /* private fields */).

Oh that makes more sense thanks.

r=me with a screenshot that the non---document-private-items appear as intended. @GrigorenkoPV do you mind posting one?

[GrigorenkoPV]

I don't quite know what the linked artifact represents because it seems like std got documented with --document-private-items there judging by CStr.inner being shown in the docs (instead of /* private fields */).

Oh that makes more sense thanks.

r=me with a screenshot that the non---document-private-items appear as intended. @GrigorenkoPV do you mind posting one?

[tgross35]

That looks like what is expected. Thanks for the update!

@bors r+ rollup

[bors]

📌 Commit b74f426 has been approved by tgross35

It is now in the queue for this repository.